Emulex Blog: Emulex Labs

I/O technology for big data has arrived, as reported by Evaluator Group

Posted October 24th, 2012 by Alex Hollingworth

John Webster, senior partner at Evaluator Group, recently wrote a Technology Insight Paper entitled, “Big Data—Maximizing the Flow,” in which he reports that many businesses in 2005 were tackling data analytics projects that required technology that was in its infant stage.1 Now, in 2012, full evolution has occurred and data analytics platforms, such as Hadoop, are allowing businesses to combine data sources and manipulate the data to maximize business benefits.

That’s Why I/O Is Strategic!

A key factor in the emerging data analytics boom is the pipe that is used to bring data from multiple sources into the data center storage where it can then be accessed for analysis and discovery. While 1Gb Ethernet (1GbE) was predominant back in 2005 when CEOs and CIOs had big plans for analyzing a lot of data, today’s data analytics projects require 10GbE to be most effective. And to get the most out of 10GbE for data analytics, the best I/O connectivity is required.

As reported in this Technology Insight Paper, bandwidth for I/O and internal cluster network is key for any Hadoop project2. Emulex offers 10GbE adapter performance improvements over 1GbE connections, where “internal networks based on 10GbE can allow a Hadoop cluster to handle increased demand for throughput without adding cluster nodes. 2” The following two charts show the difference of 1GbE throughput with 10GbE throughput.

1GbE Average Throughput10GbE Average Throughput

As companies deploy big data projects, it is helpful to gain the insights from market leaders, such as Emulex and the Evaluator Group. Read this Technology Insight Paper for all the big data tips offered by Evaluator Group. In addition, Emulex offers a Big Data Solution Site that provides resources for you as you move forward with your big data solutions. Plus, more information is available on Why I/O Is Strategic, or you can contact an Emulex representative at:

Americas: info@emulex.com
800-EMULEX3 or +1 714-885-3450
Europe: inquire.europe@emulex.com
+44 1189-772929

  1. Big Data—Maximizing the Flow, by John Webster, Evaluator Group, August 12, 2012
  2. Boosting Hadoop Performance with Emulex OneConnect 10GbE Network Adapters

Some of these products may not be available in the U.S. Please contact your supplier for more information.

The 10Gb Ethernet “Tipping Point”

Posted October 19th, 2012 by John Cedillo

When Malcolm Gladwell’s “The Tipping Point” was release in 2000, I can say with absolutely certainty that he did not intend to write a prescient piece on the lifecycle of 10Gb Ethernet (10GbE) technology. However, that does not render the application of his three “agents of change” to the current state of the technology incorrect. These agents of change – “the Law of the Few,” “the stickiness factor,” and “the power of context” – are critical elements to the tipping points of ideas, products, messages, or behaviors.

So, here we are, 10 years after the IEEE standard for 1GbE (802.3ae-2002) was established, yet widespread adoption is only just beginning. In fact, according to Dell’Oro Group (Dell’Oro Group Ethernet Switch July 2012 Forecast), we will most likely see 10GbE switch deployment surpass 1GbE in 2013, meaning 10GbE will have hit mainstream. We are standing at the precipice of the technology’s tipping point! What better time to take a brief moment and consider the “agents of change” that have led us to this point.

“The Law of the Few” – achieving critical mass in any context, is directly coupled to the involvement of individuals or organizations with very distinct skills. Gladwell delineates these skills in the following way:

  • “Connectors” are those with the ability to reach across social and business barriers to bring people together, with a combination of “curiosity, self-confidence, sociability, and energy.”
  • “Mavens” constantly accumulate knowledge on the subject and marketplace, becoming the subject matter experts, and having a ceaseless desire to share and trade for new information.
  • “Salesmen” are the persuasive and charismatic bunch with the indefinable qualities that get them all the glory. ;)

In the build-up to the current state of 10GbE adoption, there has been sufficient time for industry groups and associations to take on the role of “connectors,” technology evangelists to become “mavens,” and hardware and software innovators to become the “salesmen.”

Now, I can’t speak for other organizations, but at Emulex, I am keenly aware of our motley crew of “connectors,” “mavens,” and “salesmen.” While each skill alone is a rarity, the confluence of these skill-types is even more scarce. It’s no wonder that Emulex technology is trusted the world over.

“The stickiness factor” – delivering a message that resonates; combining what would otherwise be diametrically opposed (educational and entertaining) content in a way that people can relate…sort of like this blog?!

This is the exact reason why we have introduced our “Why I/O Is Strategic” campaign. Data centers are the lifeblood of modern business, and our products operate at the core.  While they are an easily overlooked element of the overall infrastructure, your I/O choice is strategic, and shouldn’t be left to chance.

Emulex OneConnect® Network Xceleration™ (NX) solutions take your 10GbE network and turn it into a high performance network (HPN), where low latency, better I/O analytics and predictable I/O performance is key to staying competitive. As part of Emulex’s education series “Why I/O Is Strategic,” we developed a whitepaper that describes how our OneConnect 12000-D adapters take your 10GbE to the next level. Read “Top 5 Reasons Why Emulex NX 10GbE Adapters are Strategic” to find out more.

“The power of context” – behaviors are largely influenced by our environment, and by extension, ideas, products, and messages are sensitive to the conditions and circumstances of the times in which they occur.

The technology landscape is significantly different from what it was 10 years ago:

  • In the financial vertical, High Frequency Trading (HFT) has hit the mainstream – massive profits are achieved by leveraging low latency technologies, microseconds have monetary value.

Consider this: If the low latency performance of Emulex OneConnect NX adapters with FastStack™ DBL™ can impact the revenue of a high frequency trading company by hundreds of millions of dollars in one day, think what impact even a tiny portion of that could have on your business. (read these papers to learn more: Achieve Ultra-Low Latency for High Frequency Trading (HFT) Applications and Using FastStack DBL and Sniffer10G™ in HFT platforms)

  • In the cyber/network security vertical, distributed denial of service (DDoS) attacks are a growing, and costly threat to all industries.

Consider this: In a Q1 2012 study by Neustar, DDoS attacks were found to have the potential to cost companies hundreds of thousands to millions of revenue dollars each day. What measures have you put in place to preventing this from happening to your company? Emulex OneConnect NX adapters with FastStack Sniffer10G could enable you to deploy a scalable solution to this threat.  (read this paper to learn more: Using FastStack DBL and Sniffer10G in HFT platforms and Emulex Deep Packet Inspection (DPI) Solutions)

  • In the video streaming vertical, the market is constantly being tested by the ceaselessly growing demand for media streaming to fixed and mobile devices.

Consider this: By leveraging Emulex OneConnect NX adapters with FastStack VideoPump™, you improved data center utilization by delivering more streams on fewer server while offering high quality of experience (QoE), and enabling CDN providers to further monetize data streams by freeing CPU cycles for value-add features.

The technology cycle is underway, and we’ve almost hit the tipping point. A growing percentage of the market has already started to monetize the process.  It’s worth considering taking your 10GbE infrastructure to its fullest potential and unlock the strategic value of the technology.

Solution Implementer’s Lab has new tools, papers and a new look!

Posted October 15th, 2012 by Alex Amaya

The Emulex Implementer’s Lab provides virtualization, server and network administrators with rich technical information tested and verified by Emulex Technical Marketing. This team works with partners to create and deliver tools that provide the greatest impact to Emulex end users and is your go-to for technical whitepapers and other tools to help with your deployment or management of Emulex adapters in multiple environments, including Oracle and VMware vSphere. For example:

  • Having trouble deploying your 8GB Fibre Channel (8GFC) Host Bus Adapter (HBA) with VMware in an HP environment? Read this Deployment Guide.
  • Want to reduce your Oracle batch job runtime? Read this Solution Guide.

We recently redesigned the web site and added some new features that will help you with your deployment needs and wanted to share with you the latest features, tools and resources.

As always, we have the Solution Implementers Series (SIS) technical whitepapers which provide deployment scenarios focusing on everything from installing drivers to updating firmware to supporting the appropriate OS or feature. Also included within our deployment and SIS guides are a series of application notes, which are subsets of the solution guides and provide details that may not be addressed in any of the manuals.

What’s New!

Now for what’s new. We now have a Guest Implementers page in which our customers and partners demonstrate the use of Emulex products in their blogs using real-world experiences to overcome an issue, or implement a deployment or a solution. We like to give our guests exposure to the rest of the implementers and allow them to share their knowledge, creating a more educated Emulex community. We invite you to join our Guest Implementers if you have documented your own experience with Emulex adapters!

And most exciting is our new Emulex OneCommand Manager virtual appliance (vAPP)! We’ve taken our popular OneCommand Manager application and made it into a virtual appliance. You can download it for free from the tools section of the Implementers Lab, making it easier for you to deploy in your VMware vSphere environment.

Here is your quick ‘tour’ of the redesigned Implementer’s Lab. We hope you take a tour of it soon and leverage all of the tools we’re working hard to bring you each week!

Solution Implementer's Lab

  • Home page: Highlights latest whitepaper publications – check out the new Oracle Solution Guide on how to reduce your Oracle batch run by up to an hour! Plus, a how-to application note demonstrating OneConnect® OCe12000-D with FastStack Sniffer10G for packet injection or replay – read about it!

Solution Implementer's Series

  • Implementer’s Lab Blog: Blogs written by Implementer Lab staff on the latest I/O happenings, performance testing results from inside the technical marketing lab and more
  • Tools: A collection of tools such as the free virtual appliance and tools found or created inside the technical marketing lab to ease I/O implementation. Check out the free OneCommand Manager vAPP!

Solution Implementer's Lab, Tools

  • Events: Lists webcasts, trade show schedule, presentations and more!

We would love your feedback on the redesigned web site and the tools it offers. Please send your comments to: ImplementersLab@Emulex.com.

OneCommand® Vision and Key Data Center Trends – That’s Why I/O Is Strategic!

Posted October 10th, 2012 by Erick Crowell

In our recently announced education series, “Why I/O Is Strategic,” Emulex teamed up with leading industry analysts, including Enterprise Strategy Group (ESG), IDC, Evaluator Group and 451 Research, to demonstrate the strategic importance of I/O in the data center for four key trends: virtualization, cloud computing, big data and convergence. As it happens, OneCommand Vision enhances the I/O performance and availability for each of these trends.

For each of these data center trends, if there is an I/O performance or availability issue, it usually results in lost staff time (i.e., increased operational expenditures (OPEX)) to track down the problem. Sometimes, the problem is misdiagnosed and additional servers are unnecessarily purchased (i.e., increased capital expenditures (CAPEX)). In our whitepaper, “Top 5 Reasons Why OneCommand Vision is Strategic,” we briefly explore how OneCommand Vision brings the most value to data centers large and small. For example, a huge benefit is determining in just five minutes if storage is the cause of an outage (see Figure 1).

Figure 1. Steps to rule out storage as cause of I/O issueFigure 1. Steps to rule out storage as cause of I/O issue

What OneCommand Vision does for these key data center trends, it can also do for you. Though return on investment (ROI) for this non-intrusive software application can be, and has been, immediate, you don’t need a big budget to install it. Our FREE Edition is yours for the taking – try it out!

On the importance of I/O connectivity in virtualization deployments

Posted October 4th, 2012 by Emulex Labs Team

By: Judi Uttal

It’s like alphabet soup these days when trying to understand the latest trends in I/O connectivity and how they impact virtualization deployments. NVGRE, VXLAN, SR-IOV and UMC…need help understanding all this? Emulex, a leader in I/O connectivity solutions, has asked Enterprise Strategy Group (ESG) to provide some guidance on how I/O fits with an enterprise virtualization strategy. Virtualization expert, Bob Laliberte, senior analyst at ESG, sheds insight in the Solution Brief “Why I/O is Strategic—Virtualization,” which was developed as part of Emulex’s education series “Why I/O Is Strategic.” Download this ESG Solution Brief and find out how the latest trends in I/O connectivity, like NVGRE, VXLAN, SR-IOV and UMC (which, by the way, stand for Network Virtualization using Generic Routing Encapsulation, Virtual Extensible Local Area Network, Single Root I/O Virtualization and Universal Multi-Channel™, in case you were wondering) will help you get the most out of your virtualization deployments.

VM Density Grows!

As reported in ESG’s Solution Brief, virtual machine (VM) density is growing (see Figure 1). With the growth in the number of VMs per server and highly virtualized server environments, ESG reports,“…for the first time in a long time, I/O traffic could be the bottleneck IT needs to address. ”

That’s Why I/O Is Strategic!

ESG is not the only one in the industry that knows current networking challenges for maximizing virtualization deployments make I/O a strategic decision. With mission-critical applications hosted on virtual servers, selecting the right I/O connectivity solution encompasses more than pure speed, which is why Emulex delivers “intelligence” in its adapters to support virtualization requirements, such as Quality of Service (QoS) and partitioning across different data sets.

As companies try to get the most out of their virtualization deployments, it is helpful to gain the insights from market leaders, such as Emulex and ESG. More information is available here or you can contact an Emulex representative below.

Americas: info@emulex.com
800-EMULEX3 or +1 714-885-3450

Europe: inquire.europe@emulex.com
+44 1189-772929

In addition, join our webcast with ESG: Wed, October 10 at 9a.m. PDT!

In our second webcast in a monthly series on why I/O is strategic for the data center, Bob Laliberte from ESG will shed more light on how, as virtualized environments continue to mature and become more flexible and dynamic, I/O becomes critical to a successful deployment. The webcast will also focus on the importance of assessing, designing and architecting appropriate I/O connectivity as virtualized data centers mature and become more flexible and dynamic and why I/O is important as emerging technologies enable higher virtual machine (VM) densities.

Register here and register for additional webcasts in the Why I/O Is Strategic series below:

11/14 Emulex and IDC Present Why I/O Is Strategic for the Cloud Register
12/12 Emulex and the Evaluator Group Present Why I/O Is Strategic for Big Data Register
01/16 Emulex and 451 Research Present Why I/O Is Strategic for Convergence Register

Eliminate 30 Percent Performance Penalty with New Advanced Data Integrity Solution from Emulex, EMC and Oracle

Posted October 3rd, 2012 by Sonny Singh

At OpenWorld this week, Emulex announced new second-generation LPe16000B series of 16GFC Host Bus Adapters (HBAs). Emulex 16GFC HBAs now include exclusive BlockGuard ™ data integrity offload, which supports the T10 Protection Information (T10 PI) standard, providing enhanced data protection against silent data corruption.

End-end-data integrity with T10 Protection Information (T10 PI) offload keeps data safe from silent data corruption -

Emulex is pleased to be part of the first end-to-end- T10 PI solution with Oracle and EMC. With T10 PI hardware offload on LPe16000B HBAs, data integrity checks can occur without the 30 percent IOPS performance penalty seen with the firmware-based T10 PI implementations tested by Emulex. Tests showed no IOPS performance difference with T10 PI turned on vs.T10 PI processing turned off.

Silent data corruption can go un-noticed, leading to costly system downtime, and even resulting in data loss. Until now, there was no way to detect silent data corruption until it was too late. Because silent data corruption goes undetected, replicated databases and backup copies would all contain corrupted data. When data corruption occurs, database and storage administrators must spend time recovering the database. The typical steps are:

  1. Reload the database or the affected tablespace from the last complete backup.
  2. Replay the log entries until the database or the tablespace is fully recovered. Depending on the time interval from the backup and the number of log entry replays, this step can take minutes, hours, or days.

If the backup, archive logs or online redo logs are corrupted, database recovery can be painful and may be impossible.

If the corruption happened long ago, the chances are high that recent backups would also contain the corrupted data. Unless a good backup can be found, there will be a loss of valuable data.

To prevent downtime caused by data corruptions, some businesses use duplicate copies of the same database. For example, a large data center manager uses up to six duplicate copies of their database just for quick recovery from unexplained database downtime. This increases their data center operating and hardware costs significantly. Note that these duplicate standby databases cannot be storage-based clones because a primary database corruption could be propagated to the standbys. Therefore, any standby database must be a logical copy.

The diagram below shows you how the EMC, Oracle and Emulex T10 PI end-to-end data integrity solution works:

Application to Disk Data Integrity

  1. The Oracle Automatic Storage Management (ASM) library adds integrity metadata (8 bytes) for each 512-byte sector as it is written to memory.
  2. The integrity metadata is attached to the I/O request and passed through the layers in the operating system kernel to the Emulex driver.
  3. The Emulex adapter collects the information from memory buffers, verifies the data integrity, merges the data and integrity metadata, and sends out 520-byte sectors.
  4. The EMC array firmware verifies the integrity metadata, and writes to disk.
  5. The disk drive firmware verifies the integrity metadata before committing the data to physical media.

The steps will be done in reverse when reading data.

To learn more about the EMC, Oracle and Emulex T10 PI end-to-end Data Integrity, read the whitepaper here.

Emulex will be at Oracle OpenWorld in San Francisco this week. Stop by to check out the new 16GFC adapters at the Emulex booth #943 and check out our demo of this new data integrity solution at the Oracle booth #901.

Emulex and Fusion-io Showcase Fast I/O Caching and Connectivity SAN Solution at Oracle OpenWorld

Posted October 2nd, 2012 by Emulex Labs Team

Author: Mike Jochimsen

At Oracle OpenWorld this week, Emulex announced its collaboration with Fusion-io to deliver integrated, customized I/O caching and connectivity solutions for tier 1 original equipment manufacturers (OEMs) and that at the event Fusion-io is demonstrating a solution showing how the complementary storage and networking technologies from both companies can be used in combination to help customers significantly accelerate the performance of databases, data warehouses, heavily virtualized environments and private/hybrid clouds with improved application response time and higher throughput.

The solution being demonstrated this week at Oracle OpenWorld consists of Emulex’s newly announced LightPulse® LPe16000B series of 16Gb Fibre Channel (16GFC) Host Bus Adapters (HBAs), Fusion-io ioDrive2 PCIe-based flash memory bundled with ION Data Accelerator software for caching and fast access in a shared storage FC SAN.

Customers can greatly benefit by combining these technologies in their enterprise to accelerate application response time and reduce latency for large Oracle database or data warehouses. In addition, companies are also seeing benefit for this solution in heavily virtualized environments where memory and I/O can quickly become a bottleneck.

As the diagram shows below, a customer can run Emulex’s new LPe16000B adapters in conjunction with Fusion-io ioMemory and ION Data Accelerator array used as a high performance write-through cache, interoperating with any standard Fibre Channel or iSCSI disk-based storage array on the same SAN.

Figure 1: ION and SAN Caching Interoperability

In a write-through cache, data written by the CPU to the disk array is also written into the ION storage cache. When a subsequent read operation needs that same data (also known as “hot data”), read performance is improved, because the data is already in the high-speed cache. With a write-through cache mode, all writes are sent straight through to the backing store (traditional spinning disk array) before being written to the ION cache. Since writes in the cache occur after committing them to the backing store, no user data is ever at risk. Data is as safe as the underlying backing store.¹

Also, since subsequent read operations access the ION cache, offloading these reads from the backend disk array will also make writes to the disk go up to 10x faster.²

Using Fusion-io’s ioTurbine or directCache (see below) software on the initiator server, with ION as the cache storage array, all hot data will be on the ION cache while the traditional disk array operates as the backing store.

Emulex’s LPe16000B family of 16GFC HBAs are the ideal solution with ION storage arrays for the following reasons:

  • Deliver low application latency (see Figure 2)
  • Achieve high I/O operations per second (IOPS) [see Figure 3]
  • Broad market acceptance with over 12 million installed ports
  • Certified solutions available for every major server and storage OEM
  • Deep operating system driver support

Figure 2: Emulex LPe16xxx 16GFC HBAs deliver low application latency

Figure 3: Emulex LPe16xxx 16GFC HBAs deliver high IOPS

Emulex will be at Oracle OpenWorld in San Francisco this week. Stop by to check out the new 16GFC adapters at the Emulex booth #943 and check out the joint solution demo at the Fusion-io booth #1227.


¹http://www.fusionio.com/faq/fusion-io-directcache-faq/
²http://www.fusionio.com/blog/how-much-cache-do-i-need/

Emulex Bumps Up Fibre Channel Performance by 20 Percent, Adds PCIe 3.0 and Advanced Data Integrity

Posted October 1st, 2012 by Barbara Porter

Just when you thought Fibre Channel (FC) couldn’t get any better or more reliable, Emulex has announced the LPe16000B series of 16GFC Host Bus Adapters (HBAs). This second generation 16GFC adapter delivers a new PCIe 3.0 bus and a performance boost, making it more than 20 percent faster than any other FC HBA available today. And that’s not all, check out this list of features that are only found on Emulex LPe16000B series adapters:

  • A whopping 1.2 million I/O operations per second (IOPS) on a single port¹ – the LPe16000B raises the FC performance bar yet again to support more virtual machines (VMs), larger and faster database transactions as well as delivering the fastest FC IOPS for connection to solid state disks (SSDs) and flash caching appliances. The LPe16000B features an eight processor core design, all of which can apply processing power to one port delivering exceptionally high IOPS on one port when needed. Click here to see the Demartek Performance Evaluation.
  • PCI Express 3.0 support – The LPe16000B is the only FC adapter to support PCIe 3.0, making it the perfect match to the performance capabilities of new server architectures like the Intel E5-2600 family of servers. PCIe 3.0 provides a faster I/O bus, more PCIe lanes and increased I/O bandwidth. The LPe16000B HBAs are also backward compatible with PCIe 2.0 and 4 and 8GFC infrastructures.
  • End-end-data integrity with T10 Protection Information (T10 PI) offload keeps data safe from silent data corruption – Emulex is pleased to be collaborating on the first end-to-end- T10 PI solution with Oracle and EMC. With T10 PI hardware offload on LPe16000B HBAs, data integrity checks can occur without the 30 percent IOPS performance penalty seen with the firmware-based T10 PI implementations tested by Emulex. Tests showed no IOPS performance difference with T10 PI turned on vs.T10 PI processing turned off.
  • Cloud scale reliability - Emulex HBAs are well known in the industry for providing outstanding reliability with an impressive 10 million hours mean time between failures (MTBF). With this new HBA, reliability has been upped by an additional 10 percent. This enterprise-class reliability makes a lot of sense in big deployments that simply cannot afford downtime associated with unreliable HBAs. To put this number into perspective, if a large data center has 1,000 HBAs deployed, you would expect to see approximately one failure per year.
  • In-box drivers for the latest Operating Systems (OSes) and hypervisors – The LPe16000 series is tightly integrated with new OSes and hypervisors and have in-box drivers for Windows Server 2012 Hyper-v and VMware vSphere 5.1 ensuring Emulex 16GFC adapters are supported from day one.

Emulex will be at Oracle OpenWorld in San Francisco this week. Stop by to check out the new 16GFC adapters at the Emulex booth #943 or the Fusion-io booth #1227.


¹Demartek LPe16000B Evaluation Report

An update on converged networking by 451 Research

Posted September 26th, 2012 by Emulex Labs Team

By: Judi Uttal

Since the emergence of Fibre Channel (FC)-based Storage Area Networks (SANs) in the late 1990s, enterprise IT managers have maintained two types of networks: a SAN for storage I/O traffic and a Local Area Network (LAN) for data networking traffic. The increasing adoption of blade server and server virtualization technologies has extended the demand for SANs from the traditional database tier to the application server and Web server tiers. While many IT managers continue to operate with a dual-network model across the enterprise, a number have opted for network convergence solutions realizing the benefit of reductions in overall cost structure and operational efficiency.

Fast forward to 2012. Spending on converged networking solutions, such as Converged Network Adapter (CNA) cards and Fibre Channel over Ethernet (FCoE) is increasing, as reported by TheInfoPro’s (a service of 451 Research) 2012 storage study (Figure 1). A full update on the state of converged networking can be read in 451 Research’s MIS Spotlight, When worlds combine – Storage and Data Networking Convergence, August 2012, which is part of Emulex’s education series “Why I/O Is Strategic.”

Spending Trend Chart

Figure 1. Storage Networking: 2012 vs. 2011 Spending Change of Users with a Technology In Use or In Plan

That’s Why I/O Is Strategic!

Overwhelming networking workloads, combined with application-centric networking requirements, make network adapters strategic. In fact, CIOs around the world are quickly realizing how I/O can affect their IT budgets and overall business success.*

As companies navigate converged networking solutions, it is helpful to gain the insights from market leaders, such as Emulex and 451 Research. More information is available here or you can contact an Emulex representative at:

Americas: info@emulex.com

800-EMULEX3 or +1 714-885-3450

Europe: inquire.europe@emulex.com

+44 1189-772929


*Source: InformationWeek, “Research: 2012 State of the Data Center,” 5/31/12

Time Is Money!

Posted September 19th, 2012 by Emulex Labs Team

By: Allen Ordoubadian
Time is Money

But time is relative! While we measure most of our activities in hundredths of a second, servers and networks in high frequency trading (HFT) environments measure their activities in nano-second (.0000000001 or 1 billionth of a second) time granularity. It’s simple – in HFT environments, the high performance servers, networks and trading algorithms of financial firms duke it out to see who can react the fastest – in terms of buying and selling positions – to ever changing market conditions. The reaction time to such conditions is referred to as “latency.” The firms which have the lowest overall latency typically can eke out a greater amount of financial gain. While this may sound academic, reducing latency is becoming more difficult.

Firms engaged in HFT need to consider how to decrease network-component-induced latency – in other words, the latency introduced by the network and computing infrastructure, as well as software. Many HFT designs have already co-located their trading equipment near that of the financial exchange in order to minimize signal propagation latency. Further optimization is needed, however, to reduce latency for packets traversing the switch infrastructure.

Cisco® and Emulex® have introduced an innovative low-latency solution using the Cisco Nexus® 3548 Switch and the Emulex OneConnect® Network Xceleration™ (NX) solution, leveraging the OCe12000-D high performance 10GbE network adapter and FastStack™ DBL™ network acceleration software. The testing Cisco conducted with its Nexus 3548 Switch and Emulex’s NX offering demonstrated superior performance, helping to reduce host-stack latency by up to 85 percent compared to non-optimized operations.

The chart below shows the latency results from the benchmarks Cisco conducted.

UDP Latency

To learn more, please stop by Emulex/Myricom booth at the HPC for Wall Street Conference at Roosevelt Hotel in New York City on September 19, 2012 or download the white paper at the Network Xceleration (NX) Solution with FastStack DBL page

Newer Posts»«Older Posts